Let me explain briefly what I consider the best model of how this thing we call communication works. Consider the following diagram:
At stage 1 person A is thinking about Q. Q, let us suppose, is something external to them. And so to say that they are thinking about Q is to say that they are in some particular kind of mental state (dashed-Q) that intentionally directs them at (dashed arrow) Q. The blue box represents their emotional state, in this case that they are favorably disposed towards Q. At stage 1 person B is thinking about some different object, M, which they are neutrally disposed towards. At stage 2 person A decides they wish to direct person B’s thoughts towards Q (Q may be some complex state of affairs, not necessarily a single object). To do this they make some utterance φ. φ is a completely physical phenomena; in the case of average human communication φ is usually sound waves or a some optical pattern (writing), but it of course doesn’t have to be limited to these possibilities. At stage 3, upon being affected by φ the content of person B’s mental state changes to dash-Q, and they become intentionally directed at Q. Note that how person B’s mental state changes upon being affected by φ may depend on their current mental state; which is to say that the meaning of a particular communication may be affected by context. Stages 2b and 3b detail an alternate process in which Ψ is used to communicate instead. Ψ represents emotionally charged language, not only does it put person B into a dashed-Q type mental state but is positively disposes them towards Q as well.
Given this model of communication what is language? One simple suggestion is to say that what counts as language for a particular person is anything that will reliably put them into a certain world-directed mental state upon being exposed to it. Certainly this captures what we usually think of as language, such as words, but it goes a little too far. If wheat fields regularly make me think of Kansas then that too would be considered part of the language I understand, under this definition. Now is a sense this is correct, because if you knew that wheat fields made me think of Kansas you might be able to use that fact to communicate to me about Kansas in a very non-standard way. However this isn’t what we usually mean by language. We thus restrict language to those things that dispose two or more people to adopt similar world-directed mental states (similar in that they are directed at the same thing, or at the same possible thing). And then we subtract away the trivial cases of the sensory inputs which are usually associated with those things.
Now that we understand what a language is we can move on to the more pressing question of what the meaning of particular pieces of language are. Of course I’m not looking for an explanation of particular meanings, such as dekimasu means “to be able” in Japanese, but rather of what in general meaning is. There are two schools of thought about this issue. One is that the meaning is identical with the kind of mental state it disposes the listener to be in; this is the understanding of meaning I am partial to. The other position is that meaning is to be identified solely with reference, and that other regularities in the effects it has upon the listener are to be thought of as connotations that are separate from, and thus do not have an effect on, the meaning.
Consider the theory that meaning just is reference. Now suppose that in a language L1 we had the expression “Antarctica refers to the continent at the south pole”. If meaning just is reference then if we replace expressions in L1 by those that refer to the same thing as expressions in another language, L2, and we transform the syntax appropriately, then the two sentences should mean the same thing, even though they are in different languages. We do this translation and we come up with “Antarctica refers to Antarctica” in L2. Clearly this doesn’t mean the same thing, even though our translation preserved reference (note: the point is not that we can’t translate it “correctly”, the point is that it is claimed that the incorrect translation and the correct translation have the same meaning). A partial solution to this problem is to realize that the original sentence is somewhat meta-linguistic, that what it is really saying is that “‘Antarctica’ refers to ‘the continent at the south pole’”, meaning that this is an instance of language talking about language and asserting that something about those fragments of language is similar, namely that they refer to the same thing. This helps somewhat, but our program of treating meaning as reference then runs into another problem. Specifically the original sentence refers to linguistic objects in L1; so if we translate it strictly into L2 then we are left with a sentence in L2 that makes claims about fragments of language L1. Clearly this, in one sense, has the same meaning. But in another sense it is a failure because we haven’t finished our translation. If meaning really is reference then we should be able to get an analogous sentence in L2 which makes claims about fragments of language L2 with the same meaning as the original fragments of L1. But then we run into the previous problem, where our theory of meaning asserts that “Antarctica” in L2 has the same meaning as both “Antarctica” and “the continent at the south pole” in L1.
To resolve these problems the defenders of the theory that meaning is reference invoke possible worlds to explain how “Antarctica” refers to something different from “the continent at the south pole”. I have already stated previously why I don’t think invoking possible worlds really solves the problem, so I won’t repeat myself here. However, I would like to point out that given our model of language invoking possible worlds seems completely ad-hoc. In the diagram above we have a clear and complete picture of how language works that doesn’t involve possible worlds of any kind. All that is missing from it is a label pointing to what meaning it. The only reason that possible worlds are introduced into this picture is because certain philosophers want to label the intentional arrow, or Q itself, as meaning. And to free this labeling from certain problems they introduce possible worlds, greatly complicating the intentional arrow (since now it points at a vast set of Qs), and adding a number of other new twists into the picture. This strikes me as methodologically backwards. If our model is sufficient to explain and understand communication/language then we should alter our understanding of meaning to fit our best explanation of language, not introduce more complicated explanations that don’t add to our understanding of communication or language in the least only to defend certain intuitions. And in contrast identifying meaning with the kind of mental state it tends to put the listener in suffers from none of these problems. “Antarctica” cannot be freely substituted for the phrase “the continent at the south pole”. And we don’t have to add anything more to our understanding of how language works.